Gemini 2.5 Computer Use

2025-10-07 · Publisher : Google · Domain : Artificial Intelligence

Google DeepMind releases Gemini 2.5 Computer Use in preview, a specialized model built on Gemini 2.5 Pro that drives graphical interfaces (clicks, typing, scrolling), mainly browsers.

Google DeepMind releases Gemini 2.5 Computer Use in preview, a specialized model built on Gemini 2.5 Pro that drives graphical interfaces (clicks, typing, scrolling), mainly browsers. ### What changes - New `computer_use` tool in the Gemini API, to be run in a loop (screenshot, action, new screenshot). - Optimized for browsers, promising for the mobile interface, not yet optimized for desktop operating system control. - Results superior to alternatives on several web and mobile benchmarks, with lower latency, according to Google. - Safety controls: per-step verification service outside the model, system instructions, user confirmation for risky actions. ### Specifications - Identifier: `gemini-2.5-computer-use-preview-10-2025`. - Availability: Google AI Studio, Vertex AI. ### Pricing | Tier | Input (USD / 1M tokens) | Output (USD / 1M tokens) | |---|---|---| | Prompts up to 200,000 tokens | 1.25 | 10 | | Prompts over 200,000 tokens | 2.50 | 15 | Pricing from the current official page; no free tier. ### Benchmarks The results (Online-Mind2Web via Browserbase, among others) are presented in charts, with no figures in the text of the post. ### Key takeaways A browser agent exposed through the API, with native safeguards, to be integrated with human supervision for sensitive actions.

Tags : LLM, Google, Gemini, Agents, Computer Use